LIBRA - ruQuALITY: leaderboard
LIBRA Russian QuALITY subtask: long-form multiple-choice reading comprehension over narrative passages.
Metric: Dataset Total Score (%). Source: huggingface.co. Saturation forecast: Not forecast. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 30B A3B 2507 Instruct | 85.35 | #488 |
| 2 | Llama 3.1 8B Instruct | 82.9 | #1035 |
| 3 | Phi-3.5-mini-instruct | 81.3 | #1167 |
| 4 | Qwen 3.5 35B A3B | 72.5 | #256 |
| 5 | LFM2.5-1.2B Instruct | 71.4 | #1462 |
| 6 | Qwen 3.5 122B A10B | 70.83 | #173 |
| 7 | Qwen 3.5 9B | 68.79 | #379 |
| 8 | Qwen 3.5 4B | 64.81 | #475 |
| 9 | Qwen 3.5 2B | 62 | #1148 |
| 10 | Qwen 3.5 0.8B | 58.92 | #1424 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=libra-ruquality · How It Works · Data refreshed daily, snapshot 2026-10-20.